approval-mcpapp
An MCP server that implements a multi-stage employee access request and approval workflow, enabling employees to request system access and managers/IT admins to approve or reject requests through an interactive UI.
README
Access Request & Approval Workflow — MCP App
An MCP (Model Context Protocol) server that implements a multi-stage employee access request and approval workflow. Employees request system access, and managers and IT admins review and approve or reject requests — all from within an MCP-compatible host like VS Code Copilot Chat.
Features
- Multi-stage approval pipeline — Requested → Manager Review → IT Review → Granted/Rejected
- Interactive UI panels — Rich React-based forms and dashboards served as MCP app resources
- 6 supported systems — GitHub, SAP, Production Database, Azure DevOps, Salesforce, Jira
- Persistent storage — Azure Table Storage (Azurite emulator for local dev)
Architecture
┌──────────────────────────────────────────────────┐
│ MCP Host (e.g. VS Code Copilot Chat) │
│ │
│ ┌─────────────┐ ┌──────────────┐ │
│ │ Request Form │ │ Approval │ │
│ │ (React) │ │ Panel (React)│ │
│ └──────┬───────┘ └──────┬───────┘ │
│ │ │ │
│ └────────┬───────┘ │
│ │ callServerTool │
└──────────────────┼────────────────────────────────┘
▼
┌─────────────────────────────────┐
│ MCP Server (Express) │
│ POST /mcp │
│ │
│ Tools: submit-request │
│ submit-decision │
│ get-request │
└──────────────┬──────────────────┘
│
▼
┌───────────────────┐
│ Azure Table │
│ Storage │
│ (Azurite local) │
└───────────────────┘
Prerequisites
- Node.js 22 or later
- npm 10 or later
Getting Started
1. Install dependencies
npm install
2. Start the Azurite storage emulator
In a separate terminal:
npm run start:azurite
This starts the Azure Table Storage emulator on http://127.0.0.1:10002. Data is stored in the .azurite/ directory.
3. Seed sample data (optional)
npm run seed
Loads three sample requests (REQ-001 through REQ-003) from fixtures/ so you can explore the app immediately.
4. Build and run
npm start
The MCP server starts at http://localhost:3001/mcp.
Development
Start the dev server with hot reload for both UI and backend:
npm run dev
This runs concurrently:
- UI watcher — Rebuilds HTML bundles on file changes in
ui/andsrc/ - Server watcher — Restarts the MCP server on backend file changes
Available Scripts
| Command | Description |
|---|---|
npm start |
Full build + start the server |
npm run build |
Type-check, build UI bundles, compile server TypeScript |
npm run serve |
Run the already-built server |
npm run dev |
Watch mode with hot reload |
npm run start:azurite |
Start Azurite Table Storage emulator |
npm run seed |
Seed sample data into Azurite |
MCP Tools
Frontend Tools (return interactive UI)
request-access
Opens the access request form for employees.
| Parameter | Type | Required | Description |
|---|---|---|---|
employeeName |
string | No | Pre-fill the employee name |
employeeEmail |
string | No | Pre-fill the employee email |
Example prompt: "I need to request access to GitHub"
approve-access
Opens the approval panel for managers and IT admins to review pending requests.
| Parameter | Type | Required | Description |
|---|---|---|---|
requestId |
string | No | View a specific request, or leave blank to see all pending |
Example prompt: "Show me pending access requests to approve"
Backend Tools (called by UI widgets)
| Tool | Parameters | Description |
|---|---|---|
submit-request |
employeeName, employeeEmail, system, role, justification |
Creates a new access request |
submit-decision |
requestId, decision (approve/reject), reviewer, comment |
Records an approval or rejection |
get-request |
requestId |
Fetches a single request by ID |
Workflow Stages
Each access request progresses through these stages:
┌───────────┐ ┌────────────────┐ ┌────────────┐ ┌─────────┐
│ Requested │───▶│ Manager Review │───▶│ IT Review │───▶│ Granted │
└───────────┘ └───────┬────────┘ └─────┬──────┘ └─────────┘
│ │
▼ ▼
┌──────────┐ ┌──────────┐
│ Rejected │ │ Rejected │
└──────────┘ └──────────┘
- Requested — Employee submits the form
- Manager Review — Direct manager approves or rejects
- IT Review — IT admin performs final review
- Granted — Access is provisioned
- Rejected — Request denied (can happen at either review stage)
Supported Systems and Roles
| System | Available Roles |
|---|---|
| GitHub | Read, Write, Admin |
| SAP | Finance Viewer, Finance Editor, Admin |
| Production Database | Read-Only, Read-Write, DBA |
| Azure DevOps | Reader, Contributor, Project Admin |
| Salesforce | Viewer, Editor, Admin |
| Jira | Viewer, Developer, Project Lead |
Data Model
Access Request
{
"id": "REQ-001",
"employeeName": "Alice Johnson",
"employeeEmail": "alice@contoso.com",
"system": "GitHub",
"role": "Write",
"justification": "Need write access for the frontend repo",
"status": "Manager Review",
"createdAt": "2026-03-31T10:00:00.000Z",
"updatedAt": "2026-03-31T10:00:00.000Z",
"timeline": [
{
"stage": "Requested",
"status": "completed",
"actor": "Alice Johnson",
"timestamp": "2026-03-31T10:00:00.000Z"
},
{
"stage": "Manager Review",
"status": "current",
"timestamp": "2026-03-31T10:00:00.000Z"
}
]
}
Environment Variables
| Variable | Description | Default |
|---|---|---|
PORT |
HTTP server port | 3001 |
AZURE_STORAGE_CONNECTION_STRING |
Azure Table Storage connection string | Azurite local default |
NODE_ENV |
Set to development for watch mode |
— |
Project Structure
├── main.ts # Express app entry point, /mcp endpoint
├── server.ts # MCP server definition, tool/resource registration
├── mock-data/
│ └── requests.ts # Data access layer (Azure Table Storage)
├── src/
│ ├── global.css # Shared styles
│ ├── request-form/
│ │ └── App.tsx # Employee request form UI
│ └── approval-panel/
│ └── App.tsx # Manager/admin approval panel UI
├── ui/
│ ├── request-form.html # Entry HTML for request form
│ └── approval-panel.html # Entry HTML for approval panel
├── fixtures/
│ ├── access-requests.json # Sample request data for seeding
│ └── counters.json # Counter seed data
├── scripts/
│ └── seed-data.ts # Seed script for loading fixtures
├── vite.config.ts # Vite configuration
├── tsconfig.json # TypeScript config (type-checking)
├── tsconfig.server.json # TypeScript config (server compilation)
└── package.json
Tech Stack
| Layer | Technology |
|---|---|
| MCP Framework | @modelcontextprotocol/sdk, @modelcontextprotocol/ext-apps |
| Server | Express 5, TypeScript |
| UI | React 19, Fluent UI React v9 |
| Storage | Azure Table Storage (@azure/data-tables) |
| Local Emulator | Azurite |
| Build | Vite, vite-plugin-singlefile |
| Validation | Zod |
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。